Orchestrating with MIDI (Part 1)
Modern technology can put an entire synthesized and sampled orchestra at your disposal - but could you make a listener believe he was hearing the real thing? In the first of this two-part series, Paul D Lehrman shows you how.
Electronic instruments, after working their way into so many aspects of music production, are now encountering a backlash: recording artists, commercial directors, and film and multimedia producers are demanding the sound of 'acoustic' instruments on their music tracks. Of course, they've gotten used to the cost-savings of MIDI music, and their craving for authenticity doesn't reflect in larger music budgets, so it's not quite time for trombone and hammer dulcimer players to strike up 'Happy Days Are Here Again'. But it puts the electronic musician in exactly the same predicament that the traditional instrumentalist faced ten years ago: adapt or die.
Not that it's time to throw out all your gear and learn how to play the oboe. A very large proportion of what non-musicians think are acoustic tracks are still being produced by MIDI. The people making them have mastered the art of making MIDI music sound 'real'. If you don't want to be forever typed as a 'synth composer', it's an art you should learn too.
Of course, one way to sound acoustic is to mix acoustic tracks with MIDI ones. But often that's not practical: there may be too much post-tracking editing to let you use multitrack tape (a very common situtation with film and TV work), or even if you are lucky enough to have a MIDI-with-hard-disk-audio system, the limitations on the number and length of tracks available might be too severe. So we'll concentrate here on MIDI-only compositions.
There are two basic factors that go into creating realistic-sounding music with MIDI. One is getting the right hardware. With the vast number of ROM sample players, sample+synthesis instruments, and professionally-produced sample libraries on the market today, it's actually pretty difficult not to get good hardware. (For the sake of brevity, we'll take the liberty of referring to all electronic instruments as 'synths'.) But the other part is a lot harder: learning how to play, control, and combine electronic instruments so that they sound like the real thing. Without that knowledge, you won't fool anyone but yourself.
Knowing basic principles of orchestration — and how to apply them to MIDI instruments — are important skills no matter what kind of ensemble you're trying to create: a wind band, a chamber group, a brass section, a big band, or a symphony orchestra. They're also important even when your final product is not going to be electronic, but your MIDI rig is serving as a sketch pad for a big-budget live session: if you know what works with your synths, you'll have a pretty good idea of what's going to happen when all those expensive players are in the studio.
The traditional orchestra consists of four main sections, or 'families': strings, woodwinds, brass, and percussion.
• STRINGS, which are divided into first and second violins (they're the same instruments, just playing different parts), violas, cellos, and double basses, are the workhorse of the orchestra. Each string part is played not by a single instrumentalist, but by a 'section', which can include from four to a dozen or more players.
In most classical orchestral music, strings play almost all of the time and carry the main themes of the piece. The rest of the instruments are used for colour, emphasising certain phrases, climaxes, or transitions. Early symphonic music, in fact, sometimes used just strings — witness the early works of Mozart, Haydn, and even Mendelssohn. As the Classical era moved into the Romantic, the woodwinds and brasses began to take on more of an active, independent role, but the strings still maintained their prominence.
• The WOODWIND family consists of flutes, oboes, clarinets, and bassoons. Woodwind players are soloists: one on a part. There are pairs of each instruments — for example, 1st and 2nd clarinet, 1st and 2nd bassoon — and sometimes three or four players, all playing different parts much of the time. The third or fourth players often double on auxiliary instruments like piccolo, English horn, bass clarinet, and contrabassoon, which add extra range and flexibility to the sound of the group.
French horns, believe it or not, are considered woodwinds in a symphony orchestra. Small orchestras have two horns, medium-sized ones four, and large ones eight (usually playing four parts). Saxophones are a special case: their use in the symphony was limited mainly to early 20th-century French and Russian music, and in those contexts you would usually find only one or two altos or tenors.
• The BRASSES are trumpets (two or three), trombones (two or three and sometimes a bass trombone), and tuba (usually one is plenty).
• The PERCUSSION section always consists of at least one timpani player, and often two or three others handling bass drum (which is a big, boomy thing, hit with a large, soft mallet, and is quite different from a rock kick), cymbals (a pair of them, struck against each other, creating a much more dramatic sound than a crash cymbal on a stand), snare drum, military drum (a deep snare), and/or triangle.
When you get into late Romantic and 20th century orchestras, you might also find one or two harps, a piano, a celeste, and an arsenal of fun percussion instruments: wood blocks, temple blocks, whips or slapsticks, sleigh bells, and pitched or 'mallet' percussion including xylophone, marimba, vibes, glockenspiel, and chimes (tubular bells).
Go back to the baroque era, and you'll often find a harpsichord or organ used to round out the sound of the string section (which was smaller in those days) and also to play solo lines.
Just as the real orchestra is the sum of its parts, there are many things that go into creating an electronic version of an orchestra. Each player in a real symphony knows exactly what his or her instrument is capable of doing, and how to interpret the squiggles on the page. Your task could seem daunting, because ideally you need to know everything that all those players know. But it doesn't take four years of conservatory training to learn enough to make a good orchestral sound (although it wouldn't hurt). Today's synths, thanks to their stability, sound quality, and expressiveness, take care of a lot of the basic chores. You don't need to worry about how to keep a bassoon's low F on pitch, or how to jump octaves on a tuba without splatting, how far into the bell of a French horn to put your fist to mute it, or where exactly on the violin's neck your finger needs to go to nail that high G#.
But you do need to know how to make things blend. That involves choosing the right sounds, and using them properly. The greatest string ensemble patch in the world will sound dumb next to a flat, wimpy horn section. A timpani roll in the wrong octave will wipe out your beautiful cellos. A long violin passage that is programmed in step-time, no matter how floridly and brilliantly it's written, will put the listener to sleep in seconds.
Perhaps the most important habit you can get into if you want to recreate orchestral instruments is to listen, a lot, to the real things. Listen to the way players, by themselves and in ensemble, control phrasing, note lengths, accents, the application of vibrato, and how notes decay. Try to figure out a way to make your fingers do the same thing, using velocity, aftertouch, pitchbend, and controllers. Try to input the non-note data at the same time as the notes, although if you can't, add them on a subsequent pass. If your keyboard chops are not very good, play the parts in slowly: even if you play them in at 1/10th the actual tempo, they will sound much more expressive than if you enter them in step-time.
As you assemble your orchestra, these are the main areas you should be concerned about:
• PATCH DESIGN: dynamics, envelopes, and vibrato depth and rate.
• PERFORMANCE: articulation and duration.
• ARRANGING: chord voicing and instrument doubling.
• RECORDING: panning, processing, and mixing.
We'll deal with all of these issues in various ways in this article and in Part 2 next month.
Although your goal for a project may not be to reproduce exactly the sound of a Haydn or Mozart symphony, it's helpful to know how that instrumental style is created, if the orchestral sound you do make is going to be convincing.
For starters, a string section is a very different animal from a solo string player. Many synths provide both solo and ensemble string sounds, and knowing when to use which is important.
You could create a string section out of a large number of solo instruments, with each instrument on its own MIDI channel. Playing each instrument's part individually would produce an ensemble with slight variations in timing, velocity, and controllers (vibrato) among all of the instruments. The advantage of this is that it could sound extremely realistic. The disadvantage, of course, is that it would take a long, long time, and you would probably run out of synth voices (and MIDI bandwidth) long before you were done. Therefore, string ensembles are best left to ensemble patches. If they are any good, they will already contain the slightly smeared attacks and randomized vibrato of a real string section.
But you still need to be careful, even with a great ensemble patch. Entering an orchestral string part in step-time is not a good idea, although with fast passages you can sometimes get away with it if you roughen the edges a bit, by slightly randomising the start times and durations. (Never program a solo string part in step-time.) Break the strings up into as many voices and channels as you can afford — try to put the 1st and 2nd violins on different channels — and record them all separately, even if the different sections are playing unison parts, to get as much variety into the attacks, durations, and vibrato as you can.
"A very large proportion of what non-musicians think are acoustic tracks are still being produced by MIDI. The people making them have mastered the art of making MIDI music sound 'real'"
Variations in vibrato can be very important in producing a convincing string sound. Many synth string sounds have vibrato built into them, either because they were sampled or looped that way, or from an LFO. If it's the first case, there isn't much you can do about changing it, unless you have control over the loop points. If it's an LFO, by all means change the rate slightly on the different tracks, and vary the start time of the vibrato by altering the delay parameter. If your synth lets you do these in real time, so much the better: use aftertouch to vary the rate, and scale the delay time inversely to velocity, so that softer notes start their vibrato later. A marvellous feature on some synths for livening up ensemble sounds (of all kinds, not just strings) is a 'rate-randomiser'. This automatically varies the vibrato rate from note to note, not just from track to track.
Another important factor in how real a string line sounds, particularly on a solo instrument, is how the notes are articulated. This has to do with relative durations and velocities. For example, legato quarter-notes should not be exactly one beat long: besides getting boring very fast, in most cases (unless the release portion of the envelope is quite long) it will not sound particularly legato. A little overlapping of the notes, by setting the durations to 110% of a quarter-note or so, will make the effect more convincing.
String players play up-bows (moving the bow from the tip to the handle, or frog) and down-bows (frog-to-tip), and there are subtle differences between them. Very often in fast detached passages, alternate notes will use alternating bows: down for the first eighth-note, and up for the second, and so on. A good way to achieve this is to take each second eighth-note and change its duration to about 80%, and increase its velocity by a small factor, say 5%. This will give a little 'lift' to the up-bow note, similar to the bow going up in the air and reversing directions. If the music specifically calls for two-note phrases (by putting slurs over the notes in groups of two), take this effect further by setting the second note's duration to 60% and increasing its velocity by 10%.
Designing sounds to follow the natural articulations of string instruments will help enormously. String instruments, not surprisingly, have different timbres at different dynamic levels: fortissimo playing tends to have more attack noise and high harmonics than pianissimo playing. There are also differences in the envelopes: a pp string note will normally have a much slower attack (and often release as well) than an ff note. Scaling envelope time inversely to velocity, similar to what we did earlier with vibrato delay, can add a lot of realism to string patches. Prepare to be flexible, though, because strings can play very staccato at soft volumes (particularly in spiccato passages, in which the player literally bounces the bow off the string) and very legato at loud.
If you don't have enough sampled string voices available to cover all the parts in a complex orchestration, you can often cheat successfully by filling in ensemble parts with analogue, L/A, or FM string patches. Keep the sampled voices on the top and bottom, and use the synthetic sounds for the inner voices. Synthetic strings, particularly analogue, are also very useful for fattening up the sound, if you need more body than your sampled strings provide.
Special string techniques require special patches. Pizzicato, or plucked strings, happen to be fairly difficult to synthesize (particularly in ensembles) and are best sampled. In a pinch, you can use a guitar or harp patch, if you add a little noise to the attack, make the delay very fast, use no sustain at all, and put in a small amount of ringing for the release. String harmonics, which are notated in many scores by diamond-shaped notes, are bowed, not plucked, unlike guitar harmonics. They often don't sound as written, but are an octave or a twelfth higher — check the score carefully for instructions. You can't use a normal string patch for harmonics unless you are sure it has zero vibrato. You might try creating a harmonic patch by grafting a fast, noisy string attack onto a slow-attack sine or triangle wave. The release should be slow as well.
Finally, keep in mind that cellos and basses are often notated on the same staff, but the basses transpose their part down an octave. You can sometimes get away with playing in the cello part and then simply copying it and transposing the copy down an octave, but it's often better to play the basses in separately, to give that extra variety to the parts.
One significant difference in writing for winds as opposed to strings is that you have to keep the ranges of the instruments in mind. String patches tend to take care of themselves in terms of sounding good across the pitch spectrum — if they're designed well, you don't need to know where the violin samples end and the cello samples begin. (Solo sampled strings, however, are a different story: a cello that goes too high can sound like a choking goat, while a violin that's too low might turn into a flabby rubber band.) Wind instruments have very specific ranges, and will sound odd if you force them out of those ranges. Sampled wind instruments, especially, taken too far out of range sound pretty bad. Your sampler may be able to transpose a trombone up three octaves, but you wouldn't want to listen to that sound for long, not to mention the fact that it won't resemble a trombone very much. While the sound of a string ensemble going up and down the scale is relatively uniform, the tonal characteristics of the instruments at the top and bottom of the wind families are quite different. Therefore, you have to pay close attention to which instruments play which parts, and don't let them overstep their natural boundaries.
Flutes in the bottom octave of their range tend to get buried in ensembles. Oboes, on the other hand, get strident down there. For a softer oboe tone in the low register, use an English horn. You can use a low flute as a solo instrument, and for really low parts alto or bass flute is effective. Clarinets work well right across the range, as solo or ensemble instruments, but use bass clarinet for really low parts (high bass clarinet, however, doesn't mix well at all). Bassoons can go higher than you might think in an ensemble — G above middle C is a good ceiling. Contrabassoon should only be used to reinforce the bottom of a wind ensemble, usually doubling the bassoon an octave down. A solo contrabassoon is bad enough in real life — using a sampled version all by itself is inexcusable (and I'm a former contrabassoonist). If you need solo notes that low, use a tuba, double bass, or organ pedal. Please.
French horns are considered both woodwinds and brass, in large part because they blend so well with both. Horns have an enormous range, from pedal tones almost three octaves below middle C to squeals an octave and a fourth above middle C.
The extremes of the range should be avoided however, except for loud ensemble blasts, when the sound of charging elephants is called for. In more moderate circumstances, horns make excellent mid-range and low-mid chord fillers for any kind of wind group.
Because the winds are played by only one or two players, when you make up ensembles, you have a lot of freedom to mix and match different instruments to create different orchestral colors. Doubling instruments at the same pitch or an octave apart can bring variety to the sound. Odd doublings can be very effective: piccolo and tuba three octaves apart, or an alto flute an octave below an oboe. Be careful, however, not to do anything that wouldn't work in a real orchestra: combining a low-register flute with a trombone at the top of its range won't leave much room for the flute.
In any kind of orchestral writing, but especially with winds, be careful how you lay notes across the pitch spectrum. Thick triadic chords below middle C will sound like mud if you're not careful — the lower you go, the farther the notes in a chord need to be spaced. Trombone chord voicings, therefore, tend to be open, while French horns in the middle register can be close together. Bassoons, not too loud, are okay for low chords, especially if they're moving fairly quickly. On the top end, feel free to put instruments as close together as you like — except for trumpets, unless you're specifically going for a fanfare sound.
Be aware of what each instrument is good at, and what it doesn't do so well. It's just as easy to make a synth trombone leap about quickly in its upper register as it is a synth flute, but the trombone will sound very weird while the flute will sound perfectly normal. On the other hand, long low notes on a flute will be boring and get buried easily, while they will have no trouble on a trombone or tuba.
The same basic principles of articulation we talked about earlier apply to winds as well. Tongued notes (not slurred, not staccato) should have a duration somewhere between 50% and 90% of the written value. Two-note slurs should use a ratio such as 100%/60%. Staccato notes should be 30%-50% of the duration, and legato should be exactly 100% (no overlap between adjacent notes).
In fact, the best wind legato is achieved by putting the synth in mono mode. For true legato, with no new attack, make sure the notes are at 100% of the duration, while if you want a new attack on a note, set the previous note to 95% duration. (If your synth demands that you turn on portamento to move smoothly from one note to the next, be sure to set the portamento time to 0.) For obvious reasons, this technique is not to be used when a synth is playing more than one part (both 1st and 2nd clarinets, for example) at a time.
And by the way, don't forget to breathe. Real wind players do, so be sure there's some silence between phrases, and that phrases don't go on forever. Continuous uninterrupted woodwind lines not only will sound inauthentic, they will actually cause the listener to tune out.
Vibrato, too, is as important with winds as it is with strings. Variations in vibrato between closely-related voices is crucial: nothing gives away the fact that you're using synthesizers more than an entire brass section whose vibrato is in perfect lockstep. Vibrato is also an excellent way to accentuate a voice without making it louder: add a thick LFO on a held note, and it will seem to swell even if there's no real volume change. Use vibrato in the ways that real wind players do; long notes can benefit from a vibrato that increases after the attack, changes slightly in speed over the sustain, and goes away as the note dies. Quick and short notes on winds, however, should have no vibrato at all.
Control over envelopes should also be part of the expressive mix. Even more than on strings, soft notes on winds have slower attacks, and less breath noise or 'chiff' than fast notes. If your synth has a 'sample start time' parameter, set it up so that lower velocities cause the sample to start later, thereby minimising the chiff.
Next month, we'll talk about when to use velocity and when to use volume; other kinds of real-time control; voicing chords; the percussion family; and creating a realistic orchestral image in your mix. Stay tuned.
Paul Lehrman's first orchestration with a computer was a movement from a Mendelssohn symphony, performed on an AlphaSyntauri. Although it sounded awful, he's learned a lot since then.
Part 1 (Viewing) | Part 2
Feature by Paul D. Lehrman
mu:zines is the result of thousands of hours of effort, and will require many thousands more going forward to reach our goals of getting all this content online.
If you value this resource, you can support this project - it really helps!